Mouna Torjmen Khemakhem

Toward Antifragile Cloud Computing Infrastructures

Abstract Cloud computing systems are rapidly growing in scale and complexity. They are also changing dynamically as a result of dynamic addition and removal of system components, different execution environments, common updates and upgrades, runtime repairs, mobility of devices and more. Such large-scale, complex and dynamic cloud environments are prone to failures and per- formance anomalies. Thus, dependability and resilience in cloud computing are of paramount importance to guarantee availability and reliability of services and application execution, even in the presence of large number of faulty components. Antifragility is the key to such techniques. It proposes that some systems could be strengthened by changes and faults instead of be weakened by them. In contrast to classical resilience methods, antifragile techniques aim to build systems that handle unpredictable and irregular events, while growing and getting stronger. Most of the classical resilience techniques are not sufficient to build highly available cloud infrastructures. In fact, they just resist shocks and stay the same. They should be complemented by some other aspects like learning from failure to built more elastic and stronger cloud infrastructures. This may represent the idea of building antfragile cloud systems. In this paper, we discuss the existing resilience techniques and propose a solution to design antifragile systems in cloud computing environments.

Surface modification of microfiltration ceramic membrane by fluoroalkylsilane

ABSTRACTMembrane separation performances regarding selectivity and permeate flux depend on the membrane texture and chemical composition of the material used for membrane elaboration. The surface properties of composite microfiltration membrane zirconium/mud of hydrocyclone laundries of phosphates with average pore diameter of 0.2 μm was chemically modified to change its hydrophilic feature into hydrophobic by grafting 1H,1H,2H,2H—perfluorodecyltriethoxysilane molecule (C8). Observation by scanning electron microscopy showed a sharp decrease of the grafted membrane pore size. Thermogravimetric analysis was used as a technique to identify the grafted silane groups. IR analysis allowed qualitative identification of the presence of link characteristics of incorporated silanes. The determination of the contact angle on the grafted membrane surface proved the hydrophobic character since its value increases from 25° to more than 150°, respectively before and after grafting. The membrane permeability using disti...

Semantic Annotation of Images Extracted from the Web using RDF Patterns and a Domain Ontology

Semantic annotation of web resources presents a point of interest for several research communities. The use of this technique improves the retrieval process because it allows one to pass from the traditional web to the semantic web. In this paper, we propose a new method for semantically annotating web images. The main originality of our approach lies in the use of RDF (Resource Description Framework) patterns in order to guide the annotation process with contextual factors of web images. Each pattern presents a group of information to instantiate from contextual factors related to the image to annotate. We compared the generated annotations with annotations made manually. The results we obtain are encouraging.

A Conceptual Model for Word Sense Disambiguation in Medical Image Retrieval

Word sense disambiguation (WSD) is the task of determining the meaning of an ambiguous word. It is an open problem in natural language processing because effective WSD can improve the quality of related fields such as information retrieval. Although WSD systems achieve sufficiently high levels of accuracy thanks to several technologies, it remains a challenging problem in the medical domain. In this paper, we propose a conceptual model to resolve the word sens ambiguity problem using the semantic relations between extracted concepts, through MetaMap tool and UMLS Metathesaurus. The evaluation of our disambiguation model is done through the use of information retrieval domain. Results carried out with Clef medical image retrieval 2009 show that our WSD model improves the results that are obtained by the MetaMap WSD model.

Building Contextual Implicit Links for Image Retrieval.

In context-based image retrieval, the textual information surrounding the image plays a main role in image retrieval. Although text-based approaches outperform content-based retrieval approaches, they can fail when query keywords are not matching the document content. Therefore, using only keywords in the retrieval process is not sufficient to have good results. To improve the retrieval accuracy, researchers proposed to enhance search accuracy by exploiting other contextual information such as hyperlinks that reflect a topical similarity between documents. However, hyperlinks are usually sparse and do not guarantee document content similarity (advertising and navigational hyperlinks). In addition, there are many missed links between similar documents (only few semantic links are created manually). In this paper, we propose to automatically create implicit links between images through computing the semantic similarity between the textual information surrounding those images. We studied the effectiveness of the links generated automatically in the image retrieval process. Results showed that combining different textual representations of the image is more suitable for linking similar images.

Learning to Re-rank Medical Images Using a Bayesian Network-Based Thesaurus

In this paper, we believe that representing query and images with specific medical features allows to bridge the gap between the user information need and the searched images. Queries could be classified into three categories: textual, visual and combined. We present, in this work, the list of specific medical features such as image modality and image dimensionality. We exploit these specific features in a new medical image re-ranking method based on Bayesian network. Indeed, using a learning algorithm, we construct a Bayesian network that represents the relationships among these specific features appearing in a given image collection; this network is then considered as a thesaurus (specific for that collection). The relevance of an image to a given query is obtained by means of an inference process through the Bayesian network. Finally, the images are re-ranked based on combining their initial scores and the new scores. Experiments are performed on Medical ImageCLEF datasets from 2009 to 2012 and results show that our proposed model enhances significantly the image retrieval performance compared with BM25 model.

Taking Contextual Parameters into Account and Using RDF Patterns for the Semantic Web Image Annotation

Semantic annotation of web resources presents a solution to pass from traditional web to future semantic web. Indeed, it is a process which allows to formalize extracted interpretations from web resources. In this paper, we present a new method allowing the semantic annotation of web extracted from the web. The originality of our approach lies in two points. The first point is the use of several contextual factors surrounding the image to annotate it. The second point is the construction and the instantiation of RDF (Resource Description Framework) patterns. Each pattern presents a group of information related to threaten domain. In our experimentations,we are interesting to cinema domain.

An LDA Topic Model Adaptation for Context-Based Image Retrieval

In the context-based image retrieval, the textual information surrounding the image plays a central role for ranking returned results. Although this technique outperforms content-based approaches, it may fail when the query keywords does not match the textual content of many documents containing relevant images. In addition, users are usually not experts and provide ambiguous queries that lead to heterogeneous results. To solve these problems, researchers are trying to re-rank primary results using other techniques such as query expansion, concept-based retrieval, etc. In this paper, we propose to use LDA topic model to re-rank results and therefore improve retrieval precision. We apply this model in two levels: global level represented by the whole document containing the image and local level represented by the paragraph containing an image (considered as a specific textual information for the image). Results show a significant improvement over the standard text retrieval approach by re-ranking with the LDA model applied to the local level.